Find money / currencies talk


License
MIT
Install
pip install natura==0.0.65

Documentation

natura

Build Status PyPI PyPI

natura: pun based on "paying in natura" and detecting money in natural language.

Find and convert money amounts in text.

from natura import Finder
mf = Finder()
mf.findall("500 EUR $500 and five hundred yen")
# [Money(value=500.0, currency='EUR', matches=['500', 'EUR'], base_amount=500.0, base='EUR', ...),
#  Money(value=500.0, currency='USD', matches=['$', '500'], base_amount=441.462, base='EUR', ...)
#  Money(value=500.0, currency='JPY', matches=['five', 'hundred', 'yen'], base_amount=4.404, ...)]

Takes clues to solve ambiguity:

mf.find("In Australia, 500 dollars recently became worth less")
# Money(value=500.0, currency='AUD', matches=['Australia', '500', 'dollars'], base='EUR', base_amount=336.089, ...)

Have access to the spans of the matches:

res = mf.find("In Australia, 500 dollars recently became worth less")
res.spans
# [(3, 12), (14, 17), (18, 25)]

for example, for display:

In Australia, 500 dollars recently became worth less

Install

Tested for 2.7 and 3.5:

pip install natura

Features

  • base_currency: can be changed using Finder(base_currency="USD")

  • conversion_backend: can use different websites as backends: Finder(converter=MyClass)

    Note: Default, no converter is used. To use converter https://fixer.io/, create an account and request a key.

    from natura.conversion_backends import FixerIOExchangeRate
    Finder(converter=FixerIOExchangeRate("SUPPLY_OWN_KEY"))
  • locale: create your own locales with different behavior

  • thread-safe caching: can implement different caching schedule: Finder(converter=MyClass)

    • Currently implements hourly updates and stores locally in sqlite3
  • ignore_conversions: Finder(converter=None)

  • numeric_to_money: Force numbers to become money if you know there has to be money mentioned:

Bot: "How much local money would you like to spend?"

Answer: "I would not spend more than 500"

mf.find("I would not spend more than 500", numeric_to_money=True)
# Money(value=500.0, currency='EUR', spans=[(28, 31)], matches=['500'], base='EUR', ...)
  • accurate: no false hits:
mf.find("I want to buy 500 apples.")
# None

Contributing

Yes please. There are multiple ways you could help:

  1. Implement new functionality, pull requests are very welcome
  2. Create a new locale for your language
  3. Identify cases that don't work and you think should work, pick one:
    1. Make an issue describing what doesn't work
    2. Make a pull request with an newly added test that does not pass
    3. Make a pull request with a newly added test including a FIX :)